CPU temperatures creeping up overtime after a re-paste?

[marsfarce]

Hey gang,

 

Re-pasted my y540's 9750h (and GPU) last month. Great temps initially: averages in the 70s, max CPU package mid to high 80s.

 

I kinda forgot about monitoring max temps for a while, checked HWINFO64 after playing Borderlands 3 for a few hours....94C !!! (Average was 77c). So essentially pre re-paste temperatures.

 

What's the deal? I briefly skimmed the internet and read something about not enough contact between the heatsink..air gaps..etc. Or is this really normal for a laptop CPU?

 

Additional info: Laptop always elevated behind for better airflow, undervolted by 150mV, GPU curve overclocked with OC scanner, using Arctic Mx2 TIM

[Jurrunio]

@GeneXiS_X sounds like your drying paste problem?

[genexis_x]

47 minutes ago, marsfarce said:

using Arctic Mx2 TIM

Use higher quality+higher viscosity paste.

 

Do a test: Run Aida64 system stability test, tick 'FPU' only and click start. Set performance mode in Vantage. Let me know what are the average temp for EACH core.

[marsfarce]

6 minutes ago, GeneXiS_X said:

Use higher quality+higher viscosity paste.

 

Do a test: Run Aida64 system stability test, tick 'FPU' only and click start. Set performance mode in Vantage. Let me know what are the average temp for EACH core.

Will do when I get home, thanks.

 

Vantage is always performance mode, hybrid mode disabled, Windows Power slider - Better performance.

 

 

Edited October 1, 2019 by marsfarce

[marsfarce]

Ok so I ran it with FPU only, here are the results. What do you think? It seems cores 0,2,4 are the ones reaching max temps upwards of 90c...

[marsfarce]

On 10/1/2019 at 10:56 AM, GeneXiS_X said:

Use higher quality+higher viscosity paste.

 

Do a test: Run Aida64 system stability test, tick 'FPU' only and click start. Set performance mode in Vantage. Let me know what are the average temp for EACH core.

What paste do you recommend that's really viscous?

[genexis_x]

20 hours ago, marsfarce said:

What do you think? It seems cores 0,2,4 are the ones reaching max temps upwards of 90c...

So you got the uneven core temp issue...may I know how you apply the thermal paste? (in other words, what method do you use - pea, line, spread etc)

36 minutes ago, marsfarce said:

What paste do you recommend that's really viscous?

Where are you from? Asking this bcoz not every country has all the pastes available

[marsfarce]

2 hours ago, GeneXiS_X said:

So you got the uneven core temp issue...may I know how you apply the thermal paste? (in other words, what method do you use - pea, line, spread etc)

Where are you from? Asking this bcoz not every country has all the pastes available

Interestingly after playing Overwatch for a bit last night, some of the other cores were reaching above 90 temps so I wonder is it because of uneven mounting pressure, drying out paste or low mounting pressure in general...

 

I'm from the UK.

 

I go for the pea method. 

[Falkentyne]

23 hours ago, marsfarce said:

Ok so I ran it with FPU only, here are the results. What do you think? It seems cores 0,2,4 are the ones reaching max temps upwards of 90c...

This issue even happens on LGA processors (8700K).  9900K/9700K can have core temp diff of >10C between core #1 (coolest core) and core #3 or #5 (the two hottest), assuming core 0 is first.  And all chip samples are like this since they are machine cut dies.

 

Even if you go directdie you still see the same spread, just much closer deltas. 

Generally speaking, the cores on the topside tend to be hotter (cores 0/2/4/6) than the bottom ones (1/3/5/7).

 

This happens because the die is convex.  Not the actual wafer obviously (which you can't see--check out buildzoid's old video where the silicon nitrate die is completely removed on a dead processor), but the silicone itself.  

 

This can be addressed --slightly- on LGA delidded chips by sanding down the die very slightly (like 0.1mm), which helps bring the deltas closer.  On laptops, the low static pressure heatsinks and anemic cooling just makes this difficult to address, since sanding is no longer an option.  Sometimes, if the heatsink is also convex (convex heatsink+convex die--who would've thunk it?), sanding a VERY VERY small amount off the heatsink to make it flatter can improve things, going from 400 grit to 5000 grit, then using a very thick thermal paste afterwards (like phobya nanogrease extreme, IC diamond, or another thick paste---not Kryonaut) can help address this.  On boards where the heatsink also cools the VRM's and the VRM pads are on the top side of the CPU (like where the hotter cores are), SOMETIMES reducing the thermal pad thickness by 0.5mm can help also.  Sorry that there's no one size fits all answer for you here.

[genexis_x]

1 hour ago, marsfarce said:

because of uneven mounting pressure, drying out paste or low mounting pressure in general...

Low/uneven mounting pressure. Either heatsink or CPU has bad surface

1 hour ago, marsfarce said:

I go for the pea method

Try this: use line method, but apply 2 lines (1 line on core 0/2/4 and 1 line on core 1/3/5)

1 hour ago, Falkentyne said:

This happens because the die is convex

^^this. Thanks for explaining. Well said.

[SpiralTTGL]

When apply thermal paste you want the thinnest possible layer of it hence why such a small amount is placed on the cpu and the force of the heat sink will spread it thin. Too much will in return circulate the heat acting as an insulator heating up the cpu more

[marsfarce]

2 hours ago, Falkentyne said:

Sometimes, if the heatsink is also convex (convex heatsink+convex die--who would've thunk it?), sanding a VERY VERY small amount off the heatsink to make it flatter can improve things, going from 400 grit to 5000 grit, then using a very thick thermal paste afterwards (like phobya nanogrease extreme, IC diamond, or another thick paste---not Kryonaut) can help address this.  On boards where the heatsink also cools the VRM's and the VRM pads are on the top side of the CPU (like where the hotter cores are), SOMETIMES reducing the thermal pad thickness by 0.5mm can help also.  Sorry that there's no one size fits all answer for you here.

Thanks for the reply everyone. So assuming I lightly sand the heatsink and purchase the thicker thermal paste you recommended. From your experiences, will this likely provide a longer-term solution than re-pasting every 2 months because my current, watery TIM is not doing its job? Are there any other options?

 

 

[Falkentyne]

On 10/2/2019 at 10:19 AM, SpiralTTGL said:

When apply thermal paste you want the thinnest possible layer of it hence why such a small amount is placed on the cpu and the force of the heat sink will spread it thin. Too much will in return circulate the heat acting as an insulator heating up the cpu more

This is actually not true, and Der8auer (who works for Thermal grizzly) proved it.  Excess paste just squeezes out, although this is for high pressure heatsinks like LGA/desktops.  But even on laptops it won't be more than 3-4C worse.  And that's not the issue here whatsoever.